mirror of
https://github.com/lllllllama/RigorPilot-Skills.git
synced 2026-09-14 13:43:27 +08:00
lama
118 lines
6.3 KiB
Python
118 lines
6.3 KiB
Python
#!/usr/bin/env python3
|
|
"""Exercise a skills-directory-only install without repository-level helpers.
|
|
|
|
This models the directory layout of skills installers; it does not invoke npx
|
|
or claim that a live third-party installer/network transaction was tested.
|
|
"""
|
|
from __future__ import annotations
|
|
|
|
import json
|
|
import os
|
|
import re
|
|
import shlex
|
|
import shutil
|
|
import subprocess
|
|
import sys
|
|
import tempfile
|
|
from pathlib import Path
|
|
|
|
|
|
def invoke(script: Path, args: list[str], cwd: Path) -> subprocess.CompletedProcess:
|
|
result = subprocess.run(
|
|
[sys.executable, str(script), *args], cwd=cwd,
|
|
env={**os.environ, "PYTHONIOENCODING": "utf-8", "RIGORPILOT_LESSONS": "0"},
|
|
capture_output=True, text=True, encoding="utf-8", timeout=30,
|
|
)
|
|
if result.returncode != 0:
|
|
raise AssertionError(f"installed {script.name} failed:\n{result.stdout}\n{result.stderr}")
|
|
return result
|
|
|
|
|
|
def main() -> int:
|
|
root = Path(__file__).resolve().parents[1]
|
|
with tempfile.TemporaryDirectory(prefix="rigorpilot-folder-install-") as temporary:
|
|
# Resolve platform aliases (/var vs /private/var, Windows short paths)
|
|
# before comparing this boundary with resolved installed references.
|
|
workspace = Path(temporary).resolve()
|
|
installed = workspace / "agent-home" / "skills"
|
|
shutil.copytree(root / "skills", installed,
|
|
ignore=shutil.ignore_patterns("__pycache__", "*.pyc"))
|
|
referenced_guides: set[Path] = set()
|
|
for contract in installed.glob("*/SKILL.md"):
|
|
content = contract.read_text(encoding="utf-8")
|
|
if "../../references/" in content:
|
|
raise AssertionError(f"{contract.parent.name} still requires checkout-level references")
|
|
for relative in re.findall(r"`(\.\./ai-research-reproduction/references/[^`]+\.md)`", content):
|
|
guide = (contract.parent / relative).resolve()
|
|
if not guide.is_relative_to(installed) or not guide.is_file():
|
|
raise AssertionError(f"installed {contract.parent.name} cannot resolve {relative}")
|
|
source = root / "references" / guide.name
|
|
if guide.read_bytes() != source.read_bytes():
|
|
raise AssertionError(f"installed shared guide differs from canonical source: {guide.name}")
|
|
referenced_guides.add(guide)
|
|
if not referenced_guides:
|
|
raise AssertionError("no shared guide references were exercised")
|
|
# The shared operating guide loads these sibling policies by filename;
|
|
# the variant guide resolves the advanced contract by installed skill.
|
|
guides = installed / "ai-research-reproduction/references"
|
|
for name in ("research-rigor-principles.md", "deep-learning-experiment-principles.md"):
|
|
if not (guides / name).is_file():
|
|
raise AssertionError(f"shared operating guide lost its sibling dependency: {name}")
|
|
if not (installed / "ai-research-explore/references/research-campaign-spec.md").is_file():
|
|
raise AssertionError("variant guide lost its installed campaign reference")
|
|
public_scripts = sorted(installed.glob("*/scripts/*.py"))
|
|
if not public_scripts:
|
|
raise AssertionError("no public CLI entrypoints were discovered")
|
|
for script in public_scripts:
|
|
invoke(script, ["--help"], workspace)
|
|
invoke(installed / "ai-research-reproduction/_bundled/shared/scripts/lessons_store.py",
|
|
["summarize", "--help"], workspace)
|
|
|
|
# A legacy external shared directory must not shadow a complete bundle.
|
|
stale_shared = installed.parent / "shared" / "scripts"
|
|
stale_shared.mkdir(parents=True)
|
|
shutil.copy2(root / "shared/scripts/model_adapter.py", stale_shared / "model_adapter.py")
|
|
invoke(installed / "ai-research-reproduction/scripts/run_agent.py", ["--help"], workspace)
|
|
|
|
# Execute a real, API-free fixture through the installed leaf runtime,
|
|
# then render its evidence through the installed writer fallback.
|
|
target = workspace / "target"
|
|
target.mkdir()
|
|
(target / "evaluate.py").write_text("print('accuracy=0.91')\n", encoding="utf-8")
|
|
argv = [sys.executable, "evaluate.py"]
|
|
command = subprocess.list2cmdline(argv) if os.name == "nt" else shlex.join(argv)
|
|
executed = invoke(installed / "minimal-run-and-audit/scripts/run_command.py", [
|
|
"--repo", str(target), "--command", command, "--timeout", "10",
|
|
], workspace)
|
|
payload = json.loads(executed.stdout)
|
|
if payload.get("runtime_status") != "success" or payload.get("observed_metrics", {}).get("accuracy") != 0.91:
|
|
raise AssertionError(f"installed leaf lost execution evidence: {payload}")
|
|
if "accuracy=0.91" not in Path(payload["stdout_log_path"]).read_text(encoding="utf-8"):
|
|
raise AssertionError("installed leaf did not retain the actual process stdout")
|
|
context = workspace / "context.json"
|
|
context.write_text(json.dumps({**payload, "target_repo": str(target),
|
|
"documented_command": command,
|
|
"selected_goal": "evaluation",
|
|
"goal_priority": "evaluation",
|
|
"readme_first": False,
|
|
"user_language": "en",
|
|
"result_summary": "Local fixture completed with accuracy=0.91.",
|
|
"next_action": "Inspect the retained fixture evidence."}), encoding="utf-8")
|
|
output = workspace / "repro_outputs"
|
|
invoke(installed / "minimal-run-and-audit/scripts/write_outputs.py", [
|
|
"--context-json", str(context), "--output-dir", str(output),
|
|
], workspace)
|
|
persisted = json.loads((output / "status.json").read_text(encoding="utf-8"))
|
|
if persisted.get("runtime", {}).get("status") != "success":
|
|
raise AssertionError("installed writer did not preserve verified runtime status")
|
|
print("ok: True")
|
|
print(f"public_cli_checks: {len(public_scripts)}")
|
|
print(f"shared_guide_checks: {len(referenced_guides)}")
|
|
print("checks: stale shared fallback, real execution, persisted evidence")
|
|
print("failures: 0")
|
|
return 0
|
|
|
|
|
|
if __name__ == "__main__":
|
|
raise SystemExit(main())
|